Update app.py

app.py

@@ -3,13 +3,13 @@ import numpy as np
 import tensorflow as tf
 import cv2
 from tensorflow.keras import datasets, layers, models
-import os
 
-# ========== 1. تحميل النموذج ==========
-model_path = 'mnist_cnn_model.keras'
-if os.path.exists(model_path):
-    model = tf.keras.models.load_model(model_path)
-else:
+# تحميل النموذج (إذا لم تجده، ابنه)
+try:
+    model = tf.keras.models.load_model('mnist_cnn_model.keras')
+    print("✅ تم تحميل النموذج")
+except:
+    print("⚠️ بناء نموذج جديد...")
     model = models.Sequential([
         layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),
         layers.MaxPooling2D((2, 2)),
@@ -20,48 +20,28 @@ else:
         layers.Dense(10, activation='softmax')
     ])
     model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
-    (train_images, train_labels), _ = datasets.mnist.load_data()
-    train_images = train_images.reshape((60000, 28, 28, 1)).astype('float32') / 255
-    model.fit(train_images, train_labels, epochs=3, validation_split=0.1, verbose=1)
-    model.save(model_path)
+    (x_train, y_train), _ = datasets.mnist.load_data()
+    x_train = x_train.reshape(-1, 28, 28, 1).astype('float32') / 255
+    model.fit(x_train, y_train, epochs=3, validation_split=0.1, verbose=1)
+    model.save('mnist_cnn_model.keras')
 
-# ========== 2. بيانات MNIST ==========
-(_, _), (test_images, test_labels) = datasets.mnist.load_data()
-test_images = test_images.reshape((10000, 28, 28, 1)).astype('float32') / 255
-
-# ========== 3. دوال التنبؤ والعشوائي ==========
-def predict_sketch(image):
+# دالة التنبؤ
+def predict(image):
     try:
         gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
-        resized = cv2.resize(gray, (28, 28))
-        inverted = 255 - resized
-        normalized = inverted.astype('float32') / 255.0
-        reshaped = normalized.reshape(1, 28, 28, 1)
-        pred = model.predict(reshaped, verbose=0)[0]
+        img = cv2.resize(gray, (28, 28))
+        img = (255 - img).astype('float32') / 255.0
+        img = img.reshape(1, 28, 28, 1)
+        pred = model.predict(img, verbose=0)[0]
         return {str(i): float(pred[i]) for i in range(10)}
     except Exception as e:
         return {"error": str(e)}
 
-def random_example():
-    idx = np.random.randint(0, len(test_images))
-    img = test_images[idx].reshape(28, 28)
-    img_rgb = np.stack([img] * 3, axis=2)
-    return img_rgb, f"الرقم الحقيقي: {test_labels[idx]}"
-
-# ========== 4. الواجهة (بدون launch) ==========
-with gr.Blocks(title="MNIST Recognizer") as demo:
-    gr.Markdown("# 🧠 التعرف على الأرقام")
-    with gr.Row():
-        with gr.Column():
-            sketch = gr.Sketchpad(label="ارسم هنا")
-            with gr.Row():
-                submit_btn = gr.Button("توقع", variant="primary")
-                random_btn = gr.Button("عشوائي", variant="secondary")
-            info = gr.Textbox(label="معلومات")
-        with gr.Column():
-            output = gr.Label(num_top_classes=3, label="الاحتمالات")
-    submit_btn.click(fn=predict_sketch, inputs=sketch, outputs=output)
-    random_btn.click(fn=random_example, inputs=[], outputs=[sketch, info])
-
-# ========== 5. ❌ ممنوع استخدام launch() هنا ==========
-# Spaces يتولى التشغيل عبر المتغير 'demo'
+# 🔥 تعريف demo (يجب أن يكون في المستوى العام)
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Sketchpad(label="ارسم هنا"),
+    outputs=gr.Label(num_top_classes=3),
+    title="MNIST Recognizer",
+    description="ارسم رقماً بالماوس"
+)